Add LlamaToggle command, hide completion info, and update README

- Add :LlamaToggle to enable/disable llama.vim completions without stopping server
- Disable inline timing/info display in llama.vim (show_info = 0)
- Add llama.cpp installation instructions to README

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

README.md

@@ -41,6 +41,11 @@ External Requirements:
   - If you want to write Typescript, you need `npm`
   - If you want to write Golang, you will need `go`
   - etc.
+- [llama.cpp](https://github.com/ggml-org/llama.cpp) (optional, for local AI code completion via llama.vim)
+  - macOS: `brew install llama.cpp`
+  - Other platforms: see [llama.cpp installation](https://github.com/ggml-org/llama.cpp#installation)
+  - Start the server with: `:LlamaStart` inside Neovim (uses `qwen2.5` FIM model by default)
+  - Stop the server with: `:LlamaStop` (also stops automatically when Neovim exits)
 
 > [!NOTE]
 > See [Install Recipes](#Install-Recipes) for additional Windows and Linux specific notes

init.lua

@@ -225,6 +225,10 @@ vim.keymap.set('n', '<leader>sh', '<cmd>sp<CR>', { desc = '[S]plit [H]orizontal'
 
 local llama_job_id = nil
 
+vim.api.nvim_create_user_command('LlamaToggle', function()
+  vim.cmd 'call llama#toggle()'
+end, { desc = 'Toggle llama.vim completion' })
+
 vim.api.nvim_create_user_command('LlamaStart', function()
   if llama_job_id then
     vim.notify('llama-server is already running', vim.log.levels.WARN)

lua/custom/plugins/init.lua

@@ -21,6 +21,7 @@ return {
         ring_n_chunks = 64,   -- more project context chunks
         ring_chunk_size = 64, -- smaller chunks for speed
         ring_scope = 2048,    -- wider scope around cursor
+        show_info = 0,        -- disable inline timing/info display
       }
     end,
   },